Under 50K Followers? X’s Published Algorithm Has Something You Should Know

Everybody has an opinion about the X algorithm.

“Post at exactly 8:17 AM.” “Never include links.” “Bookmarks are worth 10,000 likes.”

Cool. But what does the code actually say?

I went digging through X’s publicly available ranking code, and found something far more useful than another engagement-farming chart: the system includes special treatment for eligible posts from smaller creators.

Before anybody starts selling a $499 course about it, here’s what we can actually verify.

1. Yes, the code includes a smaller-creator boost

The published Home Mixer code contains an author_cold_start scorer. In the public implementation, eligibility includes conditions tied to the author’s follower count, whether it’s a reply or repost, how recently it was posted, and how many Home impressions it has received.

The defaults discussed in the code include:

• Up to 50,000 followers
• Not a reply or repost
• Within the first two hours
• Fewer than 200 Home impressions

This is not a guaranteed 50K-follower growth hack. It isn’t a promise that every account under the threshold gets blasted into everyone’s feeds. It’s a specific mechanism inside a much larger recommendation pipeline, and its parameters or deployment can change. It can lift one eligible post from the existing candidate pool, rather than boosting every qualifying post.

But it challenges one of the most discouraging myths on this platform: that smaller accounts have absolutely no chance of being discovered.

Takeaway: Stop waiting to be big before publishing work worth discovering.

Source: X algorithm — author cold-start scorer

2. Stop confusing model weights with points per like

You’ll see screenshots claiming that a like is worth X points, a reply is worth Y points, and a copied link is the magic ticket.

That’s not how to read a predictive ranking system.

A weight applied to a predicted probability is not the same as a fixed reward for each action someone takes. Your reader, their interests, the candidate post, the prediction model, and other ranking stages all matter.

The public code gives us clues about which outcomes the system is trying to predict. It does not give anybody a universal formula for how many impressions a like buys.

Takeaway: Write for humans who want to engage. Don’t optimize for a made-up points scoreboard.

Source: X algorithm repository

3. The post has to earn attention, not just impressions

An impression is a chance. It’s not a relationship.

I would rather have a post that somebody sends to a friend because it solved a real problem than one with a clever opening that gives them nothing after the first sentence.

The strongest posts tend to have an actual payoff:

• A shortcut someone can try in 30 seconds
• A checklist they can revisit
• A mistake you made so they don’t have to
• A working example with a source
• A genuinely surprising result they can verify

That isn’t a secret X setting. It’s a content strategy built around usefulness.

And unlike a temporary ranking parameter, it still makes sense when the algorithm changes.

4. Three algorithm myths I’d stop repeating

“Every link gets punished.” Broad, absolute claims need evidence. A link can affect user behavior and different surfaces may treat content differently, but that doesn’t establish a universal penalty on every post with a URL.

“Bookmarks have a publicly documented fixed score.” I couldn’t validate a simple universal bookmark-points rule from the published ranking material. Don’t quote invented numbers as if they’re official.

“One perfect posting time guarantees reach.” Discovery depends on more than the clock. Timing is worth testing with your audience, not worshipping as a magic minute.

The rule: separate what the source code demonstrates from what creators have merely observed.

5. My five-point preflight before hitting Post

Here’s the part worth bookmarking.

1. Does the first line give someone a reason to stop? Lead with the benefit, the surprise, or the problem. Skip throat-clearing.

2. Is there a payoff? If the hook promises a trick, show the trick. If it promises proof, bring the proof.

3. Could someone use this later? Checklists, commands, settings, comparisons, and real examples make better reference material than vague motivation.

4. Would someone send it to one specific person? “My friend needs this” is a stronger editorial standard than “maybe this will go viral.”

5. Can I defend every strong claim? Link the original source, include the date when it matters, and say what you don’t know. Nobody needs another fake algorithm cheat sheet.

The cheat sheet

FOR YOUR NEXT POST:

HOOK → PAYOFF → PROOF → SHARE VALUE → CONVERSATION

Keep it original. Make it useful. Give people something worth discussing. Measure what works for your own audience, then do more of that.

X can change its models tomorrow. That doesn’t change the value of being the person who consistently posts things other people are glad they found.

I’m going to keep digging into the published code and testing what works, rather than pretending there’s one magic number.

Save this for the next time you’re staring at the empty composer.

— Tony Simons · @tonysimons

Research note: Based on publicly available source code reviewed October 18, 2026. Public repository defaults do not necessarily describe every live production experiment or personalized feed decision.